Command Palette

Search for a command to run...

Browse

Pentagon Threatens Anthropic with $200M Cut Over Claude AI Access Demands

No image

Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei until 5:01 p.m. Friday to grant the Pentagon effectively unfettered access to its Claude AI system or face cancellation of contracts worth up to $200 million and possible blacklisting as a “supply chain risk,” in a confrontation that could reshape how the U.S. government controls frontier AI.cbsnews +1 The Pentagon has also threatened to invoke the Cold War–era Defense Production Act to compel Anthropic to tailor its systems for broader military use, an unprecedented move for AI software.theguardian

What the Pentagon Wants — and Where Anthropic Draws the Line

In a tense Tuesday meeting at the Pentagon, Hegseth told Amodei the Defense Department must be able to use Claude “for all lawful purposes,” rejecting what he has previously dismissed as “woke” usage limits on military AI.cbsnews +1 Anthropic has agreed to support U.S. national security missions but has refused two categories of use: mass domestic surveillance of Americans and fully autonomous weapons that can fire without human involvement.pbs +1

U.S. officials said other frontier AI providers — including OpenAI, Google and xAI, which also won Pentagon awards of up to $200 million last year — have accepted broader terms, increasing pressure on Anthropic, whose Claude system is currently the only model cleared for some classified military networks.washingtonpost +1 “The only reason we’re still talking to these people is we need them and we need them now,” a senior defense official told Axios, underscoring the Pentagon’s dependence on the system even as it threatens to cut ties.cbsnews

Legal and Civil-Liberties Crossroads for Military AI

The threatened use of the Defense Production Act, typically reserved for prioritizing production of physical goods like munitions or medical equipment, would push the law into new territory by compelling changes to software usage and access policies.theguardian Legal experts said simultaneously branding Anthropic a “supply chain risk” — effectively making it a pariah contractor — while forcing it to provide technology could be difficult to defend in court; “I don’t know how you square that,” said former Justice Department liaison Katie Sweeten.gizmodo

Civil-liberties advocates and AI safety researchers warned that eroding corporate guardrails against mass surveillance and autonomous weapons could set a “race to the bottom” standard across the industry, as firms weigh lucrative defense work against ethical red lines.theguardian +1 Amodei has publicly cautioned that powerful AI scanning billions of conversations could “detect pockets of disloyalty” and quash dissent, a scenario critics say looks less hypothetical as the Pentagon pushes for broad domestic data access.pbs

The Bigger Picture

The outcome of this standoff will signal how far Washington is willing to go to bend private AI labs to military priorities — and how much room companies will have to enforce their own safety rules when national security is invoked. A forced capitulation could chill safety-forward policies across the sector, while a successful resistance by Anthropic might spur Congress to clarify limits on government power over AI systems, from battlefield autonomy to surveillance of U.S. citizens.

cbsnews Axios; pbs AP; theguardian Washington Post; nytimes NPR; npr The Guardian; washingtonpost CNBC / DoD announcements; cnn New York Times; gizmodo CNN